/tools/capacity-planning
The control room ties growth, storage, traffic, fleet survival, headroom, and cost together. Scenario buttons make the trade-off visible immediately, so this walk records the changed outputs rather than just the sliders.
Live production evidence. Discovery used Codex computer control; recording used the cross-runtime capture contract so Claude can rerun the same script.
Opened the capacity-planning control room and moved to the operational scenarios.
The baseline quantifies users, replicated storage, scenario RPS, surviving instances, headroom, and monthly cost in one model.
Selected "Launch spike".
The same fleet is challenged with 2.2× demand; the scenario output rises from 87.5K to 192.5K RPS, making the capacity risk concrete.
Selected "Zone loss".
The model removes a third of the fleet while keeping peak traffic, so the learner can read the new surviving-instance and headroom consequences before touching production.
